Next Word Prediction

This repository contains machine learning models and code for predicting the next word in a sequence based on a variety of algorithms and methods, such as TF-IDF, Cosine Similarity, AdaBoost, and more. The project demonstrates how to use different techniques for text prediction tasks.

Features

• Next Word Prediction using:
• TF-IDF Multinominal Naive Bayes model / embedding with universal model encoder by Tensor Flow with LSTM model
• Training and evaluating multiple models
• Interactive Web Application using Flask
• Support for different types of word embeddings and vectorizers

Requirements

Before running the code, ensure you have the following Python libraries installed:

• Flask
• pandas
• numpy
• scikit-learn
• keras
• tensorflow
• nltk
• gensim

You can install the dependencies using the following command:

pip install -r requirements.txt

additional step for universal model encoder from Tensor Flow :

• in app.py uncomment this line

   embed = (hub.KerasLayer("https://tfhub.dev/google/universal-sentence-encoder/4"))

• make sure to make a rep in your repository named universal_model_encoder_tf
• navigate to here you find this rep and copy it in universal_model_encoder_tf

C:\\Users\\name\\AppData\\Local\\tfhub_modules

063d866c06683311b44b4992fd46fsfdsfdsf/
│
├── saved_model.pb                  
├── variables/                        
│   ├── variables.data-00000-of-00001             
│   └──  variables.index                        

• they comment the line to avoide loading the model each time you run the app

Web interface by flask framework

Note that: the corpus used is directed to concepts of AI